[Elphel-support] Cameras and processing modules for Self Driving Cars Vision System

[Elphel Support]

Marius,

About "uncompressed video" for post-processing. For post-processing we use special format of compressed images based on a simple JPEG compression, but in contrast to JPEG goal:
 "reduce bandwidth in such a way that humans will not notice the difference"
We use a different one:
"reduce bandwidth in such a way that all valid sensor data will be preserved"

Sensors output signal+noise, so the compression artifacts should just be kept lower than the natural  noise. For small-pixel sensors the dominant one is usually the shot noise.

We developed JP4 to preserve Bayer mosaic - color conversion is the most "lossy" part of the compression. It is based on the simple JPEG engine (that you can use with quality = 100% that is lossless), so it is easy to decode on any platform. It was first intended for the book scanning project ( https://en.wikipedia.org/wiki/Elphel ) , later reused for the outdoor environment scanning too.

https://wiki.elphel.com/wiki/JP4
https://community.elphel.com/jp4/jp4demo.php/

Andrey





-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://support.elphel.com/pipermail/support-list_support.elphel.com/attachments/20170630/718aa04c/attachment-0001.html>